Server-Side Request Forgery in MLflow Affects Internal Services

CVE-2026-2393

What is CVE-2026-2393?

A vulnerability in MLflow allows an authenticated attacker to exploit the _create_webhook() function, which improperly processes user-controlled URL parameters without sufficient validation. This issue enables attackers to force the MLflow backend to send HTTP requests to various endpoints, including internal services and external servers. The lack of input validation and sanitization makes the application susceptible to attacks that could lead to cloud credential theft, access to internal networks, and the exfiltration of sensitive data.

References